This thesis is part of a greater effort to use machine learning for the development of flexible and universal unresolved-scale models in large eddy simulation (LES). The novelty in the current work is that a neural network learns to predict the integral forms of the unresolved-scale terms directly without a priori assumptions on the underlying functional relationship. The contribution of this thesis is a validation of a neural-network-based unresolved-scale model for Burgers' equation which paves the way for future application to the Navier-Stokes equations.